CN103369466B - A kind of map match assists indoor orientation method - Google Patents

Abstract

A kind of map match assists indoor orientation method, relates to map match assisted location method, and solving existing indoor map matching algorithm needs to come the extra measurement data of sensor and the process of iterative algorithm search, causes the problem of computing redundancy.Arrange access point, reference point at indoor objects localizing environment, set up location fingerprint database; The CAD plane graph of floor to be positioned is utilized to generate the image array I of the floor level image of load(ing) point line model _match; Compute location result for the elements of a fix of test point; The elements of a fix are converted to pixel, judge whether have building body to stop the mid point of the n-th pixel and (n-1) individual pixel between the n-th pixel and (n-1) individual pixel then the pixel that this point of distance is nearest on Point and Line Model is searched for and by nearest pixel as revised location pixel; The location pixel of acquisition is converted to actual position coordinates, completes indoor positioning.The present invention can be widely used in indoor positioning.

Description

A kind of map match assists indoor orientation method

Technical field

The present invention relates to map match assisted location method.

Background technology

Map match assistant positioning system has become an important component part of land vehicle navigation system.High-precision numerical map can not only provide the positional information of vehicle for driver, can also be used for improving the positioning performance of integrated navigation system.At present, this technology has been widely used in outdoor vehicle navigator fix.In indoor positioning field, user adds in the randomness that diversified indoor environment is moved the difficulty that map matching technology applies in indoor environment.Meanwhile, in the complicated radio environment in indoor, due to the impact of various error, the positioning result that the fingerprint matching indoor orientation method based on WLAN obtains may depart from the normal zone of action of people.Therefore, adopt map-matching algorithm to utilize cartographic information, by the positioning result using software engineering effectively to revise WLAN, improve positioning precision.

Current, scientific research personnel has proposed a variety of indoor map matching algorithm.But, existing indoor map matching algorithm usually and some transducer conbined usage, such as, inertial sensor, gyroscope and electronic compass etc.In addition, indoor GoogleMap is utilized to adopt iterative algorithm to search for the indoor map matching algorithm of most matching result in addition.

Summary of the invention

The present invention needs to come the extra measurement data of sensor and the process of iterative algorithm search to solve existing indoor map matching algorithm, causes the problem of computing redundancy, thus provides a kind of map match to assist indoor orientation method.

A kind of map match assists indoor orientation method, and it comprises the steps:

Step one: arrange N number of access point AP at indoor objects localizing environment _i, i=1,2 ..., N, creates two-dimentional cartesian coordinate system under described localizing environment, chooses M reference point RP in described localizing environment _j, j=1,2 ..., M, the coordinate recording each reference point and W the received signal strength RSS sample gathered at each reference point place, according to M reference point RP _jcoordinate and the signal strength signal intensity RSS Sample Establishing location fingerprint database of each reference point;

Step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _match;

Described step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _matchprocess be:

Point and Line Model is loaded into after in the CAD plane graph of floor, the CAD plane graph of a load(ing) point line model can be obtained; CAD plane graph after load(ing) point line model and original CAD plane graph are saved as picture format respectively; Two width pictures are converted to respectively two bianry image matrix I _modeland I _{no_model}; In these two bianry image matrixes, the pixel in people's daily routines region is set to numerical value+1, and the pixel of building body structure and Point and Line Model is set to numerical value 0;

Obtain indoor map image array I _match:

I _{no_model}-2×(I _model⊕I _{no_model})=I _match

Wherein, ⊕ is xor operation.

Step 3: test point is set in described localizing environment, after receiving H signal strength signal intensity RSS sample at a test point place, ask for signal strength signal intensity RSS sample average and utilize fingerprint matching positioning mode k nearest neighbor algorithm KNN and weighting k nearest neighbor algorithm WKNN compute location result for the elements of a fix of test point;

Described step 3: after receiving N number of signal strength signal intensity RSS sample at a test point place, asks for signal strength signal intensity RSS sample average and utilizes fingerprint matching positioning mode k nearest neighbor algorithm KNN and weighting k nearest neighbor algorithm WKNN compute location result for the process of the elements of a fix is:

Formula is obtained according to fingerprint matching positioning mode k nearest neighbor algorithm KNN:

$\left\{\begin{array}{c}{D}_i={\left(\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{|{\mathrm{rss}}_j-{\mathrm{rss}}_{i,j}^{\mathrm{RP}}|}^q\right)}^{1/q},i=1,\·\·\·,M\\ L_r=(x_r^{\mathrm{RP}},y_r^{\mathrm{RP}}),D_r\∈\mathrm{Min}\_K\left(D_i\right),r=1,\·\·\·,K\\ (\stackrel{\‾}{x},\stackrel{\‾}{y})=\frac{1}{K}\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(x_r^{\mathrm{RP}},y_r^{\mathrm{RP}})=\frac{1}{K}\underset{r=1}{\overset{K}{\mathrm{\Σ}}}{L}_r\end{array}\right.$

Wherein, rss _jaP _isignal strength signal intensity RSS data mean value, represent at reference point RP in fingerprint database _jplace measure from AP _isignal strength signal intensity RSS data mean value; D _iit is signal strength signal intensity RSS distance; The type of q representation signal intensity RSS distance; Min_K (D _i) be the set of minimum K signal strength signal intensity RSS distance; D _r, r=1 ..., K is K minimum signal strength signal intensity RSS distance; belong to set Min_K (D _i) middle distance D _rthe coordinate of corresponding reference point;

When being assigned with weights by the position coordinates of K RP selected, described in be assigned with weights be that the inverse of signal strength signal intensity RSS distance then carries out WKNN algorithm:

$(\stackrel{\‾}{x},\stackrel{\‾}{y})=\frac{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}\left((x_r^{\mathrm{RP}},y_r^{\mathrm{RP}}){/D}_r\right)}{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(1/D_r)}=\frac{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(L_r/D_r)}{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(1/D_r)}.$

Step 4: the n-th positioning result obtained according to step 3, is converted to image array I _matchin pixel, n>=2;

Step 5: whether whether pixel described in determining step four, in building body structure, be namely I _match(x _n, y _n)=0; If be 0, this pixel, in building body structure, forwards step 7 to, otherwise forwards step 6 to;

Step 6: judge whether have building body to stop between the n-th pixel and (n-1) individual pixel, namely whether two pixels are stopped by 0 value pixel; If there is building body to stop, then forward step 7 to, otherwise forward step 8 to;

Step 7: the mid point calculating the n-th pixel and (n-1) individual pixel then the pixel that this point of distance is nearest on Point and Line Model is searched for and by nearest pixel as revised location pixel;

Step 8: the location pixel of acquisition is converted to actual position coordinates, completes indoor positioning.

The present invention adopts a kind of map match to assist indoor orientation method to realize indoor hi-Fix.The present invention makes full use of indoor map information, the positioning result of calculating is transformed to the relevant position of the image array for map match, by identifying the building body structure in this image array, irrational positioning result being adapted on Point and Line Model, more accurate positioning result can be obtained.Simultaneously relative to traditional indoor map matching algorithm, the inventive method can independently use does not need sensor measurement data, does not need iterative process.

Accompanying drawing explanation

Fig. 1 is the flow chart that a kind of map match of the present invention assists indoor orientation method;

Fig. 2 is the floor schematic diagram of specific embodiment;

Fig. 3 works as K=9 for described in specific embodiment, during H=2, is respectively k nearest neighbor algorithm KNN and positioning result schematic diagram of the present invention;

Fig. 4 works as K=9 for described in specific embodiment, during H=2, is respectively weighting k nearest neighbor algorithm WKNN and positioning result schematic diagram of the present invention.

Embodiment

Embodiment one, composition graphs 1 illustrate this embodiment.A kind of map match assists indoor orientation method, and it comprises the steps:

Step one: arrange N number of access point AP at indoor objects localizing environment _i, i=1,2 ..., N, creates two-dimentional cartesian coordinate system under described localizing environment, chooses M reference point RP in described localizing environment _j, j=1,2 ..., M, the coordinate recording each reference point and W the received signal strength RSS sample gathered at each reference point place, according to M reference point RP _jcoordinate and the signal strength signal intensity RSS Sample Establishing location fingerprint database of each reference point;

Step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _match;

Step 3: test point is set in described localizing environment, after receiving H signal strength signal intensity RSS sample at a test point place, ask for signal strength signal intensity RSS sample average and utilize fingerprint matching positioning mode k nearest neighbor algorithm KNN and weighting k nearest neighbor algorithm WKNN compute location result for the elements of a fix of test point;

Step 4: the n-th positioning result obtained according to step 3, is converted to image array I _matchin pixel, n>=2;

Step 5: whether whether pixel described in determining step four, in building body structure, be namely I _match(x _n, y _n)=0; If be 0, this pixel, in building body structure, forwards step 7 to, otherwise forwards step 6 to;

Step 6: judge whether have building body to stop between the n-th pixel and (n-1) individual pixel, namely whether two pixels are stopped by 0 value pixel; If there is building body to stop, then forward step 7 to, otherwise forward step 8 to;

Step 7: the mid point calculating the n-th pixel and (n-1) individual pixel then the pixel that this point of distance is nearest on Point and Line Model is searched for and by nearest pixel as revised location pixel;

Step 8: the location pixel of acquisition is converted to actual position coordinates, completes indoor positioning.

Embodiment two, this embodiment and embodiment one is unlike described step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _matchprocess be:

Point and Line Model is loaded into after in the CAD plane graph of floor, the CAD plane graph of a load(ing) point line model can be obtained; CAD plane graph after load(ing) point line model and original CAD plane graph are saved as picture format respectively; Two width pictures are converted to respectively two bianry image matrix I _modeland I _{no_model}; In these two bianry image matrixes, the pixel in people's daily routines region is set to numerical value+1, the pixel of building body structure and Point and Line Model is set to numerical value 0;

Obtain indoor map image array I _match:

I _{no_model}-2×(I _model⊕I _{no_model})=I _match

Wherein, ⊕ is xor operation.So, at the image array I that floor plan is new _matchin, the pixel of Point and Line Model is replaced by numerical value-1, and the pixel of people zone of action is replaced by numerical value+1, and the pixel of building body structure is replaced by numerical value 0.

Embodiment three, this embodiment and embodiment one or two are unlike described step 3: in described localizing environment, arrange test point, after receiving H signal strength signal intensity RSS sample at a test point place, ask for signal strength signal intensity RSS sample average and utilize fingerprint matching positioning mode k nearest neighbor algorithm KNN and weighting k nearest neighbor algorithm WKNN compute location result for the process of the elements of a fix of test point is:

Formula is obtained according to fingerprint matching positioning mode k nearest neighbor algorithm KNN:

$\left\{\begin{array}{c}{D}_i={\left(\underset{j=1}{\overset{N}{\mathrm{\Σ}}}{|{\mathrm{rss}}_j-{\mathrm{rss}}_{i,j}^{\mathrm{RP}}|}^q\right)}^{1/q},i=1,\·\·\·,M\\ L_r=(x_r^{\mathrm{RP}},y_r^{\mathrm{RP}}),D_r\∈\mathrm{Min}\_K\left(D_i\right),r=1,\·\·\·,K\\ (\stackrel{\‾}{x},\stackrel{\‾}{y})=\frac{1}{K}\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(x_r^{\mathrm{RP}},y_r^{\mathrm{RP}})=\frac{1}{K}\underset{r=1}{\overset{K}{\mathrm{\Σ}}}{L}_r\end{array}\right.$

Wherein, rss _jaP _isignal strength signal intensity RSS data mean value, represent at reference point RP in fingerprint database _jplace measure from AP _isignal strength signal intensity RSS data mean value; D _iit is signal strength signal intensity RSS distance; The type of q representation signal intensity RSS distance; Min_K (D _i) be the set of minimum K signal strength signal intensity RSS distance; D _r, r=1 ..., K is K minimum signal strength signal intensity RSS distance; r=1 ..., K belongs to set Min_K (D _i) middle distance D _rthe coordinate of corresponding reference point;

When being assigned with weights by the position coordinates of K RP selected, described in be assigned with weights be that the inverse of signal strength signal intensity RSS distance then carries out WKNN algorithm:

$(\stackrel{\‾}{x},\stackrel{\‾}{y})=\frac{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}\left((x_r^{\mathrm{RP}},y_r^{\mathrm{RP}}){/D}_r\right)}{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(1/D_r)}=\frac{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(L_r/D_r)}{\underset{r=1}{\overset{K}{\mathrm{\Σ}}}(1/D_r)}.$

Embodiment four, this embodiment and embodiment three, unlike described q=2, namely chooses Euclidean distance.

Embodiment five, this embodiment and embodiment one or four are positive integer unlike described N, M, and W, H are positive integer.

Specific embodiment:

Composition graphs 2-4 illustrates this specific embodiment.The validity of the method is tested under the laboratory experiment environment shown in Fig. 2.Wherein, 9 LinksysWAP54GAP are arranged in the laboratory experiment environment of 24.9m × 66.4m.Experiment path is from the A point room to the B point in the corridor of 3 meters wide.Experiment utilizes an Asus A8F notebook computer image data.It is equipped with IntelPRO/Wireless3945ABG wireless network card and RSS data acquisition software NetStumbler, sampling rate 2 RSS data samples per second.In off-line phase, select in a room to select 67 reference points in the corridor of 24 reference points and 3 meters wide, with collections in 150 seconds totally 300 RSS samples in each reference point.At on-line stage, choose 65 test points along experiment path A altogether to B, spacing is 0.6m.Consider quantity and the layout of reference point, arrange K=9, when H=2, q=2, calculate traditional KNN and WKNN algorithm and assist the positioning performance after indoor orientation method with application map match, the mean error of experimental result is more as shown in table 1.

Table 1 algorithm performance compares

Algorithm	Mean error (m)
		KNN	2.73
WKNN	2.70
		KNN+Map	2.45
WKNN+Map	2.47

Shown in Fig. 3 and Fig. 4, at turning, corridor, because Structure in Complex Structure is larger on radio propagation impact, so the positioning result of KNN algorithm and WKNN algorithm is poor, have a lot of positioning result to have passed through body of wall, this is impossible in practice.Apply after map match proposed by the invention assists indoor orientation method, positioning performance obviously improves, and has been adapted on the Point and Line Model proposed by irrational positioning result, obtains the elements of a fix more accurately.The mean error of positioning result also drops to 2.45 meters and 2.47 meters respectively from 2.73 of KNN and WKNN algorithm meters and 2.70 meters.Therefore, above-mentioned experiment has convincingly demonstrated value and the validity that map match proposed by the invention assists indoor orientation method.

Claims (2)

1. map match assists an indoor orientation method, it is characterized in that it comprises the steps:

Step one: arrange N number of access point AP at indoor objects localizing environment _i, i=1,2 ..., N, creates two-dimentional cartesian coordinate system under described localizing environment, chooses M reference point RP in described localizing environment _j, j=1,2 ..., M, the coordinate recording each reference point and W the received signal strength RSS sample gathered at each reference point place, according to M reference point RP _jcoordinate and the signal strength signal intensity RSS Sample Establishing location fingerprint database of each reference point;

Step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _match;

Step 3: test point is set in described localizing environment, after receiving H signal strength signal intensity RSS sample at a test point place, ask for signal strength signal intensity RSS sample average and utilize fingerprint matching positioning mode k nearest neighbor algorithm KNN or weighting k nearest neighbor algorithm WKNN compute location result for the elements of a fix of test point;

Step 4: the n-th positioning result obtained according to step 3, is converted to image array I _matchin pixel, n>=2;

Step 5: whether whether pixel described in determining step four, in building body structure, be namely I _match(x _n, y _n)=0; If be 0, this pixel, in building body structure, forwards step 7 to, otherwise forwards step 6 to;

Step 6: judge whether have building body to stop between the n-th pixel and (n-1) individual pixel, namely whether two pixels are stopped by 0 value pixel; If there is building body to stop, then forward step 7 to, otherwise forward step 8 to;

Step 7: the mid point calculating the n-th pixel and (n-1) individual pixel then the pixel P that this point of distance is nearest on Point and Line Model is searched for _n'=(x _n', y _n'), and by nearest pixel P _n' as revised location pixel;

Step 8: the revised location pixel obtained is converted to actual position coordinates, completes indoor positioning.

2. a kind of map match according to claim 1 assists indoor orientation method, it is characterized in that described step 2: utilize the CAD plane graph of floor to be positioned to generate the image array I of the floor level image of load(ing) point line model _matchprocess be:

Point and Line Model is loaded into after in the CAD plane graph of floor, the CAD plane graph of a load(ing) point line model can be obtained; CAD plane graph after load(ing) point line model and original CAD plane graph are saved as picture format respectively; Two width pictures are converted to respectively two bianry image matrix I _modeland I _{no_model}; In these two bianry image matrixes, the pixel in people's daily routines region is set to numerical value+1, the pixel of building body structure and Point and Line Model is set to numerical value 0; Obtain indoor map image array I _match:

I _{no_model}-2×(I _model⊕I _{no_model})＝I _match

Wherein, ⊕ is xor operation.